21474
J. Phys. Chem. C 2010, 114, 21474–21481
A Theoretical Study of CO2 Anions on Anatase (101) Surface Haiying He,† Peter Zapol,*,†,‡ and Larry A. Curtiss†,‡,§ Materials Science DiVision, Chemical Sciences and Engineering DiVision, and Center for Nanoscale Materials, Argonne National Laboratory, Argonne, Illinois 60439 ReceiVed: July 15, 2010; ReVised Manuscript ReceiVed: September 24, 2010
Binding configurations of CO2 and CO2- on perfect and oxygen-deficient anatase (101) surfaces were explored using first-principles calculations on both cluster and periodic models. The solvent effect was taken into account via the polarizable continuum model. Analysis of molecular orbitals, charge, and spin density distributions was used to help identify the radical anion CO2- adsorbed on the surface. On defect-free surfaces, it is found to bind as a bridging bidentate configuration with both oxygens coordinating to the 5-fold Ti ions. Analysis of vibrational frequencies provides a specific signature of the CO2 anion to distinguish it from other species in experiments. The reduction potential of adsorbed CO2 on a (101) surface is lower by 0.24 V than the reduction potential of a CO2 molecule, both in aqueous solution, due to the formation of hybridized orbitals, which facilitates charge transfer to CO2. The reduced (101) surface of TiO2 is much more favorable for CO2 binding with accompanying charge transfer to CO2. I. Introduction Photochemical CO2 reduction into hydrocarbon fuels by harnessing solar energy offers potential economic and environmental benefits.1-3 Efficient recycling of green-house CO2 gas and fixing light energy in chemical bonds is an appealing alternative energy technology. Since discovery that a suspension of semiconductor powders (such as TiO2, SiC, GaP, etc.) could photocatalyze the reduction of CO2 with water as a reductant to produce formic acid, formaldehyde, methanol, and methane under UV light,4 much effort has been devoted to finding suitable sensitizers for visible-light adsorption and to increasing the activity of the reaction. The photochemical reduction of CO2 has also been observed in water vapor experiments in recent studies.5 In the photocatalytic process, anatase TiO2, although not absorbing more than 5% of solar light, is considered to be an outstanding candidate due to efficient separation of electron-hole pairs and chemical stability. Often a cocatalyst such as Pt metal is used to facilitate reduction of the adsorbed CO2 by photoexcited electrons on the semiconductor surface. Despite many studies on high-efficiency and high-selectivity photocatalysts over the past few decades, the conversion efficiency of CO2 into useful hydrocarbons is still much too low for technological applications.3 Further progress requires better understanding of the reaction mechanisms. The conversion of CO2 to fuels is an eight-electron process and is made up of a sequence of steps involving electron and proton transfer, C-O bond breaking, and C-H bond formation. Since CO2 is a rather inert molecule with a positive electron affinity (0.6 ( 0.2 eV)6 indicating that the process CO2 + e f CO2- is not favorable, the initial activation of CO2 on a photocatalyst surface is the key issue. It is widely believed that the first step in CO2 conversion is formation of the anion radical CO2- on TiO2 surface by exciting an electron to its lowest-unoccupied molecular orbital (LUMO).2 * To whom correspondence should be addressed: Email: zapol@ anl.gov. Phone: 630-252-6085. Fax: 630-252-9555. † Materials Science Division. ‡ Chemical Sciences and Engineering Division. § Center for Nanoscale Materials.
Nonetheless, there are only a few experiments that have reported observation of CO2- on a pure TiO2 surface under illumination using vibrational spectroscopic techniques.7 These studies provide no details on the bond lengths or angles, and the assignments of vibrational frequencies are tentative.8 There is a lack of theoretical studies in this respect as well. Previous theoretical studies were concerned with CO2 on TiO2 in ground9,10 and excited states,11 CO2 and CO2- on other oxides,12-14 and CO2- in the gas phase.15 The existence of the CO2- radical anion on the photoactive TiO2 surface remains an unsettled question and a critical issue in our current understanding of the photoreduction process and mechanism of CO2 conversion. In this study, we address this question based on results of first-principles calculations of the anatase (101) surfacesthe most abundant surface of anatase.16 Here, we have explored a variety of possible binding configurations of neutral and charged CO2 on the anatase (101) surface in terms of geometries, energies, charge distribution, and vibrational frequencies. Detailed analysis of molecular orbitals and spin densities is further used to help to identify the radical anion CO2-. We have also calculated the binding geometries of CO2 on oxygen-deficient anatase (101) surfaces, which is of interest because of the presence of oxygen vacancies on TiO2 surfaces in experiments. II. Computational Model and Method Photoexcitation of electron-hole pairs in titania results in charge separation, with electrons available to participate in the CO2 reduction. Here, we assume that the dynamics of charge separation is much faster than electron transfer to CO2; therefore the electrons available for reduction are thermalized to the bottom of TiO2 conduction band (CB), which is about 3.8 eV below the vacuum level.17 Photoexcited electrons are modeled by introduction of an extra electron into the system, which populates the conduction band of TiO2. The holes which compensate these electrons and interact with them electrostatically are modeled in three possible ways: (i) completely screened and having no influence (model I); (ii) positive background charge (model II); (iii) localized positive charge in the vicinity,
10.1021/jp106579b 2010 American Chemical Society Published on Web 11/10/2010
CO2 Anions on Anatase (101) Surface
J. Phys. Chem. C, Vol. 114, No. 49, 2010 21475
modeled by a proton (model III). The activation of CO2 involving one-electron transfer is accompanied by a large deformation from a linear structure to a bent structure with an OCO angle of about 138°. In the gas phase, the bent structure is a metastable state of CO2 anion, and CO2 has a positive electron affinity suggesting that its LUMO is above the vacuum level. Thus, electron transfer from TiO2 to CO2 molecule is likely to happen in the binding configurations of bent CO2 within the Born-Oppenheimer approximation. The electron distribution is determined in a self-consistent field procedure, which results in electron transfer to CO2 when it is favorable. Taken together, a large electron-hole separation and a barrier on the adiabatic energy surface of CO2 anion justify our use of ground state methods to compute properties of a problem otherwise requiring explicit treatment18 of electronic excited states. Both clusters and periodic slabs simulating the TiO2 surfaces were used in our calculations to accommodate the three models I, II, and III that we have proposed above in treating the Coulomb effects of electron-hole interactions. In model I, a cluster of Ti7O27H26, representing the anatase (101) surface, was adopted. The cluster was cleaved from an anatase (101) surface16 with added H atoms to terminate oxygen dangling bonds. These hydroxyl groups were kept fixed in the structure relaxations. An extra electron is introduced in the system to model the photoexcited electron, whereas the hole is completely screened and not considered. Previous studies19 using DFT+U to study adsorption on reduced rutile (110) surfaces found that the location of the initial unpaired electron has little effect on the energies of the surfaces and that the final states of the surfaces with an adsorbate in case of charge transfer is the same. Anatase (101) surface features two types of undercoordinated atoms: 5-fold Ti atoms and 2-fold O atoms. Several orientations of adsorbed CO2 molecule at different binding sites on the cluster were explored. All these calculations were performed within the framework of the DFT with the B3LYP functional20,21 as implemented in the Gaussian03 program package.22 The 6-31+G(2df,p) basis set was used for the CO2 molecule and five surface O atoms near the 5-fold-Ti reaction center (see Supporting Information for details), and the LanL2DZ basis set was used for Ti, H, and the remaining O atoms. We have included the solvation effects using the polarizable continuum model (PCM).23 In this model, the water solvent is represented by a homogeneous continuum medium having a dielectric constant of 78.39, which is polarized by the solute placed in a cavity built in the bulk of water. We report only the electrostatic contribution to the solvation energy, since the cavitation and dispersion terms are dominated by the cluster choice. Also note that these terms almost cancel out in calculations of binding energies in the solution. The reduction potential of CO2/CO2(in the form of a single molecule or an adsorbate on TiO2 surfaces) with respect to the normal hydrogen electrode (NHE) was calculated according to
Ereduct(NHE) ) (Eaq(CO2) - Eaq(CO2 ))/F - 4.50V
(2) where 4.50 V is the absolute reduction potential of the NHE24 and F is the Faraday constant. Vibrational frequencies for different surface binding configurations in the gas phase were calculated and scaled by a factor of 0.9652.25 Natural bond orbital (NBO) charge analysis26 was carried out for all the Gaussian calculations. The binding configurations of CO2 on TiO2 in the presence of an oxygen vacancy were also studied. The creation of an oxygen vacancy
by removing one oxygen atom from the bridging site leads to two extra electrons in the vacancy region or effectively two Ti3+ ions, which in turn might reduce CO2 to CO2-. In the periodic slab systems, we have constructed the anatase (101) surface using a 2 × 1 supercell along [010] and [101- ] directions consisting of 6 TiO2 trilayers with 4 Ti atoms per layer. We have fixed the atoms in the bottom trilayer to their bulk positions. A vacuum layer of about 11 Å was placed along the z direction. All the calculations were done using plane wave basis sets and PAW-PBE27 pseudopotentials implemented in the VASP program.28 A Monkhorst-Pack grid of 2 × 2 × 1 was used to sample the first Brillouin zone of k space. The total energy is converged to 10-5 eV, and the force on each atom is relaxed to 0.03 eV/Å. To account for the presence of an electron from photoexcitation, either an extra electron with a positive compensating background (model II) or one hydrogen atom adsorbed on the surface bridging O atom (model III) is introduced to populate the TiO2 conduction band. In the latter case, the electron from the hydrogen atom redistributes to the bottom of CB of TiO2 within self-consistent calculations leaving behind a proton mimicking a localized hole in terms of its Coulomb potential as long as it does not participate in reactions. In contrast, we consider the positive charge to be completely screened in our cluster calculations. Bader charge analysis29 was done to analyze charge populations in the periodic calculations. III. Results and Discussion CO2 and CO2- in the Gas Phase and Water Solvent. The calculated geometries, electron affinities, and vibrational frequencies of the isolated CO2 and CO2- molecules are compared to previous experiments or high level theoretical calculations as shown in Table 1. CO2 molecule has a positive electron affinity. Our calculated electron affinity of CO2 (0.43 eV) is within the experimental error of the measured value of 0.6 ( 0.2 eV,6 while it is about 0.2 eV lower than the higher-level calculations based on the coupled-cluster method (see Table 1).15,30 Our calculations of CO2/CO2- in aqueous solution result in an electrostatic solvation energy of 0.11 and 2.82 eV for CO2 and CO2-, respectively. They are in reasonable agreement with the reported experimental free energy of solvation values of 0.0931 eV and 3.21 ( 0.0832 eV for CO2 and CO2-, respectively. The agreement indicates that the electrostatic interaction is the dominant contribution to these solvation free energies in water. It leads to a CO2 reduction potential of -2.22 V relative to NHE, in a reasonable agreement with the experimental value of -1.93 V.32 Adding and optimizing three explicit water molecules along with the PCM results in a calculated value of -2.09 V. By consideration of the calculated gas-phase vibrational frequencies reported in Table 1, it is interesting that the shift for the antisymmetric stretching frequency from CO2 to CO2is about 700 cm-1, which is in agreement with the observed downward shift. In water solvent, the CO2- frequencies ν1 (symmetric stretching) and ν2 (bending) both increase by ∼50 cm-1 compared to the gas phase, while ν3 (antisymmetric stretching) is reduced by ∼130 cm-1. CO2 and CO2- on Perfect Anatase (101) Surfaces. Adsorption energies, charge, and spin distributions for CO2 adsorption on neutral and negatively charged TiO2 surfaces are reported in Tables 2 and 3. Optimized geometries for adsorbed CO2 on charged surface in the cluster model are given in Figure 1, and optimized geometries for adsorbed CO2 on a charge neutral surface as well as all geometries for periodic models are given in Supporting Information.
21476
J. Phys. Chem. C, Vol. 114, No. 49, 2010
He et al.
TABLE 1: Calculated Geometries, Energies, and Vibrational Frequencies for CO2 and CO2- in the Gas Phase and in Water Solvent Using the PCM Model Performed at the B3LYP/6-31+G(2df,p) Levela gas
water CO2-
CO2 this work energy (eV) R(C-O) (Å) ∠OCO (deg) freq (cm-1)
ν2 ν1 ν3
0.0 1.16 180.0 664 1325 2327
other calc
exptl
0.0 1.161b 1.1614c 180.0 687c 1359c 2417c
this work
0.0 1.162d 180.0 667e 1286e 2349e
0.43 1.23 137.9 639 1174 1699
CO2-
CO2
other calc a
b
exptl
this work
this work
0.6 ( 0.2 1.25g 127 ( 8g 134h 849g 1424g 1671g 1658I
-0.11 1.162 180.0 652 1326 2287
-2.39 1.237 134.9 696 1223 1570
e
0.63 0.66 1.24b 1.2301c 138b 137.93c 679c 1222c 1769c
a Energies are relative to the energy of an isolated CO2 molecule. The total energy of the system in water solvent includes the electrostatic energy of solvation. Vibrational frequencies were scaled by a factor of 0.9652.25 b EOM-CCSD/Aug-cc-pVTZ level. See ref 15. c CCSD(T)/6-311+G(3df) level. See ref 30. Note that the vibrational frequencies were not scaled. d See ref 46. e See ref 47. f See ref 6. g See ref 48. h See ref 49. I See ref 50.
TABLE 2: Calculated Adsorption Energies (in eV) of CO2 on Neutral Anatase (101) Surface Using Cluster and Periodic Models and on Negatively Charged Anatase (101) Surfaces Using the Cluster Model (I) and the Periodic Model (II) and H Reduced Anatase (101) Surface Using the Periodic Model (III)c state
neutral
anion
model config
cluster
periodic
other
I
II
III
A1 A2 B1 B2
-0.34(0.18) -0.41(-0.19)
-0.20 -0.14
-0.34b -0.72c
-0.35(-0.28) -0.35(-0.67) 0.33(0.54) 0.21(0.23)
-0.07 -0.23 0.72 -0.03
-0.21 -0.21 0.50 -0.06
0.22(0.94)
0.06
a Energies are relative to the sum of the energy of neutral or negatively charged or H reduced TiO2 surface and the energy of a CO2 molecule. The adsorption energies in water solvent are included in parentheses. b See ref 11. c See ref 10.
TABLE 3: Calculated Charges and Spin Distributions of CO2 Adsorbed on Neutral Anatase (101) Surface Using Cluster and Periodic Models and on Negatively Charged Anatase (101) Surface Using Cluster Model (I) and Periodic Model (II) and H Reduced Anatase (101) Surface Using Periodic Model (III)a neutral charge model config cluster periodic A1 A2 B1 B2
0.10 -0.23
0.00 -0.04
-0.18
-0.11
anion charge I 0.09 -0.31 -0.45 -0.16
II
spin III
I
II
-0.03 0.01 0.03 0.00 -0.11 -0.08 0.01 0.00 -0.65 -0.60 0.87 0.81 -0.18 -0.12 0.00 0.00
III 0.00 0.00 0.78 0.00
a The values of charge or spin are the sum of the atomic charges on CO2 either by NBO (for the cluster model) or by Bader (for the periodic models) charge analysis.
In the case of neutral CO2 adsorbed on anatase (101) surface, three bound configurations A1, A2, and B2 were found. The A1 and A2 configurations both have positive binding energies and are energetically competitive. A1 represents a nearly linear configuration of CO2 vertically adsorbed on top of a 5-fold Ti atom (similar to the A1 structure in Figure 1a for a charged surface), while A2 is bent and has one side laying on a Ti(5fold)-O(2-fold) bond on the surface (similar to the A2 structure in Figure 1b for a charged surface), which has also been identified in a previous theoretical study of neutral CO2 adsorption on TiO2 surfaces by Indrakanti et al.10 They report10 a binding energy of 0.72 eV for the A2 configuration, which is higher than our result of 0.41 eV in model I. The difference is due to their use of a relatively small cluster model Ti2O9H10. As a result, a terminating H atom in their cluster forms a hydrogen bond with CO2.10 The binding energy of the linear adsorption configuration A1 obtained from the cluster model is in very good agreement with the value of 0.34 eV reported from a previous theoretical study of CO2 (linear configuration)
interacting with excited stoichiometric anatase (101) surfaces.11 In our calculations, the binding energies predicted by the periodic slab model (0.20 and 0.14 eV for A1 and A2, respectively) are lower than those of the cluster model (0.34 and 0.41 eV for A1 and A2, respectively) due to the use a different functional, PBE, and different basis set in the periodic calculations. The generalized gradient approximation PBE functional tends to underestimate the binding energy. This is also reflected in the smaller charge transfer and larger bonding distances in the corresponding periodic model (see Table 3 and Supporting Information). A higher energy metastable configuration B2 is also found with two O atoms bridging two 5-fold Ti atoms and C atom pointing downward and forming a C · · · O bond with the 3-fold O atom on the surface (see, Figure 1d). The formation of this configuration is endothermic. The four distinct binding configurations (A1, A2, B1, and B2) have been identified in the presence of an extra electron on the TiO2 surface from the cluster calculations (model I, as illustrated in Figure 1) and periodic calculations (models II and III). As shown in Tables 2 and 3, the three models produce rather similar trends in the energy ordering, charge and spin distributions of the four configurations. Three of these configurations (A1, A2, and B2) are very similar structurally and energetically to their analogous configurations of neutral CO2 on anatase (101) surface. In models I and III, the linear adsorption A1 configuration (Figure 1a) and the bent configuration A2 (Figure 1b) have the same energy. The adsorption energy and the distance (especially in the case of the cluster model) for A1 between the Ti and the nearest O of the CO2 molecule (see Figure 1a and Figure S3a, for instance) are almost the same as in the neutral case, apparently unaffected by the presence of the extra charge on the surface. Deskins et al. studied the role of excess electrons originating from oxygen vacancies, bridging row hydroxyls, and interstitial Ti species in the surface chemistry using DFT+U.19 Almost the same adsorption energy
CO2 Anions on Anatase (101) Surface
J. Phys. Chem. C, Vol. 114, No. 49, 2010 21477
Figure 1. Configurations of CO2 on perfect anatase (101) surfaces with an extra negative charge in the system (model I). Symbols for atoms: Ti in green, O in red, H in white, C in blue. Distances are in Å, and angles are in degrees.
of CO2, among other species, on rutile (110) surfaces has been found regardless of whether an extra electron has been present. In model II, however, the binding energy of A1 reduces to 0.07 eV, making CO2 barely adsorbed, which is consistent with the long distance of 2.84 Å (see Supporting Information) between the 5-fold Ti atom and the anchoring O atom of CO2 molecule. The binding energy of CO2 in the A2 configuration is 0.16 eV higher than that of the A1 configuration in model II. The largely decreased binding energy of linear CO2 in model II is probably due to the small negative charge of 0.03e on CO2, which results in a decrease of CO2 binding energy due to electrostatic repulsion with a negative charge on the surface. The large deformation of CO2 in the A2 configuration caused by bonding of O of CO2 molecule to a surface Ti and of C to a surface O as compared to the linear structure of CO2 molecule in the gas phase and in A1 configuration indicates that there is an energy barrier to achieve this configuration. The barrier between A1 and A2 is found to be 0.5 eV by a nudged elastic band calculation33 in model III. The B2 configuration is found to be higher in energy than the A1 configuration. The relatively large difference in binding energy of the B2 configuration in the three models is a result due to both the difference in physical model and the sensitivity of this configuration to the adopted exchangecorrelation functional forms. For instance, using the same exchange-correlation functional, PBE, for cluster calculations, the energy difference between B2 and A1 is decreased from 0.56 to 0.30 eV, which is in closer agreement with the slab model using the same functional. It is worth pointing out that all of the three configurations A1, A2, and B2 are energetically very similar to the neutral systems. The charge populations of atoms in CO2 are also about the same, suggesting that the additional electron is distributed on Ti atoms on the surface. In the charged state, we found another structure, B1 (Figure 1c) that is not present in the neutral system. Structure B1 has a bridging bidentate binding configuration with two O atoms bridging two 5-fold Ti atoms and the C atom pointing upward without forming any bond with the surface atoms. The average Ti-O(CO) distances are 2.10, 2.18, and 2.18 Å for models I, II and III, respectively, well within the chemical bonding range
Figure 2. Spin densities of different binding configurations of CO2 on perfect anatase (101) surfaces with an extra negative charge. Symbols for atoms: Ti in green, O in red, H in white, C in blue.
of Ti-O. The two relatively strong Ti · · · O bonds hold CO2 in place with bent OCO angles of 135.4°, 136.4°, and 137.1° for models I, II, and III, respectively. These value are close to the angle of CO2- in the gas phase of 138°. Nonetheless, B1 is found to be the most unfavorable one in total energy, higher by 0.68, 0.79, and 0.71 eV than the linear adsorption configuration A1 for models I, II, and III, respectively. Note that B1 is not stable at all in the neutral system. The extra charge is shown to be critical to stabilize this binding configuration. Note that this bidentate binding of CO2 is not sensitive to the position of the adsorbed H atom on the surface in model III. The NBO and Bader charge analyses further indicate that in the B1 configuration, CO2 is negatively charged. And most strikingly it shows that there is an unpaired electron (spin) located mostly on the central C atom, indicating that C is indeed reduced in this case, while it is not the case for all the other configurations (Figure 2). Although the A2 and B2 configurations also have negative charge on the CO2 part, B1 is the only one that has net charge transfer to the C atom. The net charges on A2 and B2 are about the same in the anionic case as in the neutral case, which indicates the observed net charges are mainly because of chemical bonding between the adsorbates and the surface
21478
J. Phys. Chem. C, Vol. 114, No. 49, 2010
without forming an anion radical of CO2-. This leads to the conclusion that a CO2- radical anion on perfect anatase (101) surfaces takes the form of B1. The B1 configuration is similar in terms of geometry of CO2 moieties to larger molecules adsorbed on TiO2 surfaces that form bridging bidentate. For example, previous studies34-38 of formate and acetate show that the CO2 moiety with a negative charge adopts a similar bridging bidentate configuration. By implicitly considering the solvent effect using PCM for cluster models and taking into account the electrostatic part of the solvation energy, we have calculated the binding energies in aqueous solution for all configurations in both neutral and anion cases. The values are given in Table 2. It is found that in water all of the binding energies with exception of A2 decrease as compared to the gas phase. Overall, the drop in the binding energy is much larger in the neutral system than in the anionic system, with the decreases in the neutral system being 0.22 eV (A2), 0.52 eV (A1), and 0.72 eV (B2). This reflects the largely decreased polarization of the solvent surrounding the [TiO2 · · · CO2] complex as compared to the polarization of the solvent due to the TiO2 cluster alone in the absence of CO2 adsorbate, since the solvation effect of the nonpolar CO2 molecule is negligibly small. In the presence of an excess electron in the system (the anion case), however, the drop in the binding energy is relatively small, typically within 0.2 eV. It is notable that in the case of the anionic A2 configuration (a carbonate-like structure) the binding energy increases by 0.32 eV in solution, indicating its largely enhanced stability in water solution. This is likely due to the favorable combination of contributions from the asymmetric binding geometry of CO2 on the TiO2 surface, the partial charge transfer from TiO2 to CO2, and the relatively large dipole moment of the complex. The A2 configuration is 0.39 eV lower in energy than the A1 configuration in solution. The energies of the anionic B1 and B2 configurations are 0.82 and 0.51 eV higher than that of A1, respectively. And interestingly, we have noticed that the release of CO2- from the B1 binding configuration into water solution lowers the total energy of the system by 0.12 eV due to the large solvation energy of CO2- anion. In most cases, the effect of solvent is destabilization of adsorbed CO2 and CO2-; therefore, it seems easier to reduce CO2 on the surface in the absence of the solvent. The reduction potential of CO2/CO2on TiO2 (101) surface is calculated to be -1.98 V relative to the NHE as compared to the value of -2.22 V in the case without TiO2 substrate. This change in the reduction potential indicates that the reduction of CO2 on TiO2 surface is easier than without TiO2 in the solution. To investigate dependence of adsorption energies on surface coverage, we have performed additional calculations for the neutral and anionic structures A1 and B1 with two adsorbed CO2 molecules per 2 × 1 surface cell and one adsorbed CO2 molecule per 4 × 2 surface cell in addition to the results for one CO2 molecule per 2 × 1 unit cell discussed above. The results (given in Supproting Information) indicate that the A1 neutral configuration has a very weak dependence on coverage with adsorption energy variations from -0.20 to -0.24 eV per CO2. A slight increase in binding at lower coverage indicates slight repulsive interactions among adsorbed molecules at higher coverage. For the B1 anion configuration, the situation is reversed with higher coverage being energetically more favorable due to increase in interactions of negatively charged CO2 molecules with positive countercharges at the surface. In addition, we have investigated coadsorption of the B1 and one other configuration (A1, A2, and B2) in 2 × 1 surface cell with
He et al.
Figure 3. The SOMOs of CO2 on anatase (101) surfaces in anionic (a) B1 and (b) A1 configurations.
the conclusion that the B1 configuration can be significantly stabilized by coadsorption with A2 and B2. Electronic Properties of CO2- Anion Radical on Anatase Surface. The initial activation of CO2 to generate CO2- requires one-electron transfer from the surface to the adsorbed molecule. In the gas phase, the negatively charged molecule CO2- was found to be metastable and 0.43 eV higher in energy than the neutral state. On the anatase (101) surface, the favored accommodation of the bent CO2 anion on a pair of its unsaturated Ti sites stabilizes this charged molecule. The energy of the CO2radical on TiO2 (B1 configuration) is 1.76 eV lower in energy than the sum of the energies of these two species at large separation. It indicates that the electron transfer to CO2 adsorbed on the surface is much more energetically favorable than the electron transfer to CO2 in the gas phase. The bidentate adsorption of CO2 on TiO2 forms the hybridized LUMO composed of CO2 orbitals and the 3d orbitals of Ti atoms, located in the semiconductor gap. The significant lowering of the LUMO of CO2 allows electron transfer from the titanium 3d band to CO2. This is confirmed by the molecular orbital analysis, as shown in Figure 3. In the case of the B1 configuration, the singly occupied molecular orbital (SOMO) is mainly distributed over CO2 molecule and two binding Ti atoms, featured by a highly hybridized electronic state; while in the case of A1, SOMO has negligible contribution from CO2. Therefore, it is clear from these results that the photocatalyst semiconductor not only generates electron-hole (e-h) pairs via the photoexcitation but also can provide an active site for catalysis by adsorbing the reactant and lowering its LUMO. The active interplay between the catalyst surface and reactants, intermediates and products will affect the overall reaction rates and selectivity significantly. This opens the opportunity of design of new catalysts through surface engineering. Vibrational Frequencies of Absorbed CO2 and CO2- on Perfect Anatase (101) Surfaces. IR and Raman spectra are often used to identify different species on surfaces. Therefore, we have calculated the vibrational frequencies of absorbed CO2 and CO2-, which are given in Table 4. The results are from cluster model calculations. The calculated antisymmetric stretching frequency of the adsorbed CO2 (linear configuration A1) on TiO2 surface agrees with the experimental value of 2340 cm-1 for linear CO2 on TiO2 rutile (110) surface.39 The relative shift of frequencies of CO2 after adsorption on the surface is also in good agreement with experimental observations,8 where a linearly coordinated CO2 on MgO, Fe2O3, TiO2, ZrO2, and Al2O3 was characterized by a shift up of the ν3 vibrational mode by 13-31 cm-1 and the shift down of the 2ν2 Fermi resonance doublet by 66 cm-1. The slight shift down of the ν1 mode (∼11 cm-1), however, is contrasted by a small shift up in our case. The lowering of its symmetry after CO2 being absorbed on the surface in the nearly linear configuration (A1) leads to lifting up the degeneracy of
CO2 Anions on Anatase (101) Surface
J. Phys. Chem. C, Vol. 114, No. 49, 2010 21479
TABLE 4: Calculated Vibrational Frequencies (in cm-1) of Bending (ν2), Symmetric Stretching (ν1), and Asymmetric Stretching (ν3) Modes of CO2/CO2- in the Gas Phase, Water Solvent, and on Perfect Anatase (101) Surfacesa anatase (101) surface neutral anion
a
mode
gas
water
A1
ν2 ν1 ν3 ν2 ν1 ν3
664 1325 2327 639 1174 1699
652 1326 2287 696 1223 1570
640, 632 1332 2346 635, 626 1329 2341
B1
B2
A2
692 1265 1638
802 1254 1706 817 1267 1678
817 1086 1776 822 1101 1740
Vibrational frequencies were scaled by a factor of 0.9652.25
TABLE 5: Calculated Adsorption Energies (eV), NBO Charges, and Vibrational Frequenciesa of CO2 Adsorbed on Anatase (101) Surfaces with Oxygen Vacanciesb VO-A1 charge on CO2
freq (cm-1)
VO-A2
VO-B1
VO-B2
Eads
-1.09
-0.36
0.08
-0.97
C O O total ν2 ν1 ν3
0.64 -0.58 -0.69 -0.63 724 921b 1766b
0.52 -0.75 -0.68 -0.91 765 910b 1194b
0.83 -0.63 -0.70 -0.50 705 1236 1689
0.60 -0.56 -0.61 -0.57 744 1133 1704
a Vibrational frequencies were scaled by a factor of 0.9652.25 Due to stronger bonding of one of the CO2 oxygen atoms with the defective surface (filling the oxygen vacancy site on the surface) in the case of the VO-A1 and VO-A2 configurations, the assignments of vibrational modes ν1 and ν3 are for two C-O stretching modes other than the symmetric and asymmetric stretching modes. b
ν2 mode, as shown in Table 4 for A1 configurations in both neutral and anionic systems. IR spectra were previously used to identify bent CO2 species on TiO2 surfaces.7,8 The values reported by Rasko´ et al.,7 1640 and 1219 cm-1 bands, match quite well with our B1 bands of 1638 and 1265 cm-1. Our calculated values for B2 and A2 agree with the bands of two species observed in the experiments of Ramis et al.,8 namely, the bands that they have assigned to sideon bonded CO2 on cationic sites and C-bonded species,
respectively. But, as is apparent from our discussion in the previous section, these two species are not CO2- radicals because of the small net negative charge and negligible spin density on the CO2 moiety. The strong shift in the range of 600-700 cm-1 in the CO2 antisymmetric stretch frequency (ν3), however, is common to all the three bent structures B1, A2, and B2. B1 is distinct in maintaining a frequency of ∼700 cm-1 for the bending mode (ν2), which is much higher (around 820 cm-1) for A2 and B2 configurations. Therefore, the combined features of these two modes (ν3 and ν2) may serve as a unique signature to distinguish the B1 configuration from other bent configurations, such as A2 and B2. The vibrational frequencies show fairly similar values for both neutral and anionic systems in the same configuration (with notable exception of B1), suggesting the extra charge of the system is largely distributed on TiO2 and has a negligible effect on the adsorbate. CO2 Adsorption on Oxygen-Deficient Anatase (101) Surfaces. Oxygen-deficient TiO2 has also been of much interest in photocatalysis40-42 owing to the fact that a lower valence state of Ti3+ is introduced into the system by the creation of oxygen vacancies (VO). Oxygen vacancies in the materials can be generated by annealing in vacuum, UV radiation, ion sputtering, plasma treating, and sol-gel techniques.43 In our calculations using the cluster model, a bridging O atom is removed to generate an O vacancy, resulting in the two electrons redistributing over Ti 3d states.19 This electron-rich defective system may help to form and stabilize CO2 anions. The calculated CO2 binding energies and NBO charges are listed in Table 5. Four distinct binding configurations were also found for CO2 adsorbed on defective anatase (101) surfaces, derived from the binding configurations on the perfect anatase (101) surface. The optimized structures are shown in Figure 4, labeled as VO-A1, VO-A2, VO-B1, and VO-B2. The most stable configuration is obtained from linear CO2 adsorbed at an oxygen vacancy by filling the vacancy and bending toward the 5-fold Ti with an ∠OCO angle of 128°. The calculated binding energy (1.09 eV) is much higher than the binding energy of the most stable CO2 configuration on perfect TiO2 (101) surface. It is in close analogy to the adsorption of CO on perfect anatase (101) surface except that C is solely bonded onto the 5-fold Ti in the latter.44 The dissociated CO bonding configuration is about 0.1 eV higher
Figure 4. Calculated CO2 adsorption geometries on the cluster model of anatase (101) surfaces with oxygen vacancies. Symbols for atoms: Ti in green, O of TiO2 in red, O of CO2 in orange, H in white, C in blue. Distances are in Å, and angles are in degrees.
21480
J. Phys. Chem. C, Vol. 114, No. 49, 2010
in binding energy than the VO-A1 configuration. The VO-B2 is slightly less stable than the VO-A1 configuration. In the VO-B2 configuration C-Ti bond is formed with C pointing downward and filling the vacancy. The chelate-type structure VO-A2 is much less energetically favorable than the VO-B2 structure. The VO-B1 configuration, which is a close analogue of the B1 configuration on perfect (101) surface (but along a different orientation), is unfavorable by 0.08 eV. This is 1.17 eV lower in binding energy than the most stable configuration VO-A1. It is interesting to note that this energy difference is higher than that of B1/A1 on negatively charged perfect anatase (101) surfaces. This structure is in a triplet state with one spin mainly on CO2 and the other on a 5-fold Ti cation (see Supporting Information). Therefore, we can see that the presence of point defects favors bent CO2 conformations and thus helps to stabilize negatively charged CO2- on the surface. Calculated vibrational frequencies for CO2 adsorbed on the surface with defects (see Table 5) exhibit strong shifts with respect to those on both neutral and negatively charged surfaces. This is not surprising given large changes in bonding and geometry discussed above. The reaction routes following the activation of CO2 also depend on the adsorption geometry of CO2- on TiO2 surface. This adds to the complexity of reactions and may affect the selectivity of products. It is worth noting that the binding energies of H2O molecule and O2 molecule are 0.64 and 2.78 eV higher than that of CO2 molecule on the vacancy site. Therefore, in the presence of H2O or O2, CO2 is less likely to adsorb on the oxygen vacancy site.45 The recovery of vacancy sites in a catalytic process may be another critical issue for turnover on such active sites. IV. Conclusions We have performed cluster and periodic first-principles studies of CO2 negative ion adsorption on perfect and defective anatase TiO2 surfaces in gas and water environments. We have identified three distinct adsorption configurations of CO2 adsorbed on the surface and one additional configuration for a CO2 negative ion. The latter is a bridging bidentate configuration that has both oxygens of CO2 coordinating to surface 5-fold Ti ions. Spin distribution indicates that unpaired spin is mostly localized on the carbon atom. Calculated vibrational frequencies suggest the specific signature to identify adsorbed CO2 negative ion in IR experiments. Calculations using a solvation model indicate destabilization of CO2 binding by solvent both in neutral and charged systems. This suggests that it should be easier to observe CO2 negative ions adsorbed on anatase from the gas phase. As expected, the reduced surface of TiO2 with oxygen vacancies, is much more favorable for CO2 binding with accompanying charge transfer to CO2. These results clarify the structure and properties of CO2- on anatase, widely believed to be important species in photo- and electrocatalysis, and open the opportunity of design of new catalysts through surface engineering. Acknowledgment. Work, including use of the Center for Nanoscale Materials, is supported by the U.S. Department of Energy under Contract DE-AC0206CH11357. We acknowledge grants of computer time from EMSL, a national scientific user facility located at Pacific Northwest National Laboratory and the ANL Laboratory Computing Resource Center. Supporting Information Available: Geometries of CO2 adsorbed on neutral anatase (101) surfaces in both the cluster and the periodic slab models; geometries of CO2 adsorbed on
He et al. negatively charged anatase (101) surfaces in models II and III; test results for the cluster model using larger basis sets and a larger cluster size; molecular orbital analysis of CO2 adsorbed on anatase (101) surfaces with an oxygen vacancy. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Roy, S. C.; Varghese, O. K.; Paulose, M.; Grimes, C. A. ACS Nano 2010, 4, 1259. (2) Centi, G.; Perathoner, S. Catal. Today 2009, 148, 191. (3) Indrakanti, V. P.; Kubicki, J. D.; Schobert, H. H. Energy EnViron. Sci. 2009, 2, 745. (4) Inoue, T.; Fujishima, A.; Konishi, S.; Honda, K. Nature 1979, 277, 637. (5) Varghese, O. K.; Paulose, M.; LaTempa, T. J.; Grimes, C. A. Nano Lett. 2010, 10, 750. (6) Compton, R. N.; Reinhardt, P. W.; Cooper, C. D. J. Chem. Phys. 1975, 63, 3821. (7) Rasko, J.; Solymosi, F. J. Phys. Chem. 1994, 98, 7147. (8) Ramis, G.; Busca, G.; Lorenzelli, V. Mater. Chem. Phys. 1991, 29, 425. (9) Markovits, A.; Fahmi, A.; Minot, C. THEOCHEM 1996, 371, 219. (10) Indrakanti, V. P.; Kubicki, J. D.; Schobert, H. H. Energy Fuels 2008, 22, 2611. (11) Indrakanti, V. P.; Schobert, H. H.; Kubicki, J. D. Energy Fuels 2009, 23, 5247. (12) Hammami, R.; Dhouib, A.; Fernandez, S.; Minot, C. Catal. Today 2008, 139, 227. (13) Preda, G.; Pacchioni, G.; Chiesa, M.; Giamello, E. J. Phys. Chem. C 2008, 112, 19568. (14) Pan, Y. X.; Liu, C. J.; Wiltowski, T. S.; Ge, Q. F. Catal. Today 2009, 147, 68. (15) Sommerfeld, T.; Meyer, H. D.; Cederbaum, L. S. Phys. Chem. Chem. Phys. 2004, 6, 42. (16) Lazzeri, M.; Vittadini, A.; Selloni, A. Phys. ReV. B 2001, 63, 155409. (17) Redmond, G.; Fitzmaurice, D. J. Phys. Chem. 1993, 97, 1426. (18) Duncan, W. R.; Prezhdo, O. V. Annu. ReV. Phys. Chem. 2007, 58, 143. (19) Deskins, N. A.; Rousseau, R.; Dupuis, M. J. Phys. Chem. C 2010, 114, 5891. (20) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (21) Lee, C. T.; Yang, W. T.; Parr, R. G. Phys. ReV. B 1988, 37, 785. (22) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03; Gaussian, Inc.: Wallingford, CT, 2004. (23) Tomasi, J.; Mennucci, B.; Cammi, R. Chem. ReV. 2005, 105, 2999. (24) Pearson, R. G. J. Am. Chem. Soc. 1986, 108, 6109. (25) Merrick, J. P.; Moran, D.; Radom, L. J. Phys. Chem. A 2007, 111, 11683. (26) Reed, A. E.; Curtiss, L. A.; Weinhold, F. Chem. ReV. 1988, 88, 899. (27) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (28) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169. (29) Sanville, E.; Kenny, S. D.; Smith, R.; Henkelman, G. J. Comput. Chem. 2007, 28, 899. (30) Gutsev, G. L.; Bartlett, R. J.; Compton, R. N. J. Chem. Phys. 1998, 108, 6756. (31) Wagman, D. D.; Evans, W. H.; Parker, V. B.; Schumm, R. H.; Halow, I.; Bailey, S. M.; Churney, K. L.; Nuttall, R. L. J. Phys. Chem. Ref. Data 1982, 11, 1. (32) Koppenol, W. H.; Rush, J. D. J. Phys. Chem. 1987, 91, 4429. (33) Henkelman, G.; Jonsson, H. J. Chem. Phys. 2000, 113, 9978. (34) Persson, P.; Ojama¨e, L. Chem. Phys. Lett. 2000, 321, 302. (35) Vittadini, A.; Selloni, A.; Rotzinger, F. P.; Gratzel, M. J. Phys. Chem. B 2000, 104, 1300.
CO2 Anions on Anatase (101) Surface (36) Foster, A. S.; Nieminen, R. M. J. Chem. Phys. 2004, 121, 9039. (37) Ojama¨e, L.; Aulin, C.; Pedersen, H.; Ka¨ll, P. O. J. Colloid Interface Sci. 2006, 296, 71. (38) Ka¨ckell, P.; Terakura, K. Surf. Sci. 2000, 461, 191. (39) Henderson, M. A. Surf. Sci. 1998, 400, 203. (40) Diebold, U. Surf. Sci. Rep. 2003, 48, 53. (41) Ihara, T.; Miyoshi, M.; Ando, M.; Sugihara, S.; Iriyama, Y. J. Mater. Sci. 2001, 36, 4201. (42) Bonapasta, A. A.; Filippone, F. Surf. Sci. 2005, 577, 59. (43) Suriye, K.; Jongsomjit, B.; Satayaprasert, C.; Praserthdam, P. Appl. Surf. Sci. 2008, 255, 2759. (44) Scaranto, J.; Giorgianni, S. THEOCHEM 2008, 858, 72.
J. Phys. Chem. C, Vol. 114, No. 49, 2010 21481 (45) Green, J.; Carter, E.; Murphy, D. M. Chem. Phys. Lett. 2009, 477, 340. (46) Herzberg, G. Molecular Spectra and Molecular Structure III. Electronic Spectra and Electronic Structure of Polyatomic Molecules: Van Nostrand Reinhold: New York, 1966. (47) Kazansky, V. B.; Borovkov, V. Y.; Serykh, A. I.; Bulow, M. Phys. Chem. Chem. Phys. 1999, 1, 3701. (48) Hartman, K. O.; Hisatsune, I. C. J. Chem. Phys. 1966, 44, 1913. (49) Ovenall, D. W.; Whiffen, D. H. Mol. Phys. 1961, 4, 135. (50) Jacox, M. E.; Thompson, W. E. J. Chem. Phys. 1989, 91, 1410.
JP106579B